I tested the 50 most active company MCP servers — 27 scored a median 85 with recurring gaps
abhiramreddykalluru · reddit · 2026-10-31
Using QAmcp, the author's free open-source tester (runs only tools labeled read-only, changing nothing), he tested the 50 most frequently updated company MCP servers from the official registry:
- 23 of 50 required a login key and couldn't be tested
- The other 27 scored a median of 85/100 (lowest 71, highest 96) — no disasters, but the same gaps kept appearing:
- 16 of 27 had warnings on tool descriptions — what an AI reads to decide what to call
- 13 had speed warnings
- 5 had no tools that could be safely run (mostly because nothing was labeled read-only), so AI clients can't tell what's safe either
- 11 had more than 40 tools; one had 754
Biggest surprise: even servers shipping several times a week still have these gaps. Tester: https://qamcp.dev/
More from coding & agent
- Selling AI agents to paper-bound Balkan SMBs: how do you close non-tech traditional businesses? — Then-Occasion18 · 2026-11-01
- Watch an AI agent run fully autonomous in YOLO mode: session recording — umeshonai · 2026-11-01
- Claude autonomously directs an AI music video: 80 renders, ~12 GPU-hours, just 8% of a $20 plan — Cheap_Credit_3957 · 2026-11-01
- DeepSeek Details Always-On Agent Design With External Watchdog Session for Progress Monitoring — teortaxesTex · 2026-11-01
- Gemini 4 roasts Google's 3D harness: blind coordinate writing and how to fix it — Ryoiki-Tokuiten · 2026-11-01
- Weekly: Anthropic's S1 filing, labs claim models refute Exponential Time Hypothesis — njyx · 2026-11-01